REVIEW
How Do MicroRNAs Regulate
Gene Expression?
Richard J. Jackson* and Nancy Standart*
(Published 2 January 2007)
Several thousand human genes, amounting to about
one-third of the whole genome, are potential targets
for regulation by the several hundred microRNAs
(miRNAs) encoded in the genome. The regulation
occurs posttranscriptionally and involves the ~21-
nucleotide miRNA interacting with a target site in the
mRNA that generally has imperfect complementarity
to the miRNA. The target sites are almost invariably
in the 3′-untranslated region of the messenger RNA
(mRNA), often in multiple copies. Metazoan miRNAs
were previously thought to down-regulate protein
expression by inhibiting target mRNA translation at
some stage after the translation initiation step, with-
out much effect on mRNA abundance. However,
recent studies have questioned these suppositions.
With some targets, an increase in the rate of mRNA
degradation by the normal decay pathway con-
tributes to the decrease in protein expression.
miRNAs can also inhibit translation initiation, specif-
ically the function of the cap-binding initiation factor,
eIF4E. Repressed target mRNAs as well as miRNAs
themselves accumulate in cytoplasmic foci known
as P-bodies, where many enzymes involved in mRNA
degradation are concentrated. However, P-bodies
may also serve as repositories for the temporary and
reversible storage of untranslated mRNA, and reduc-
ing the expression (knockdown) of several distinct
P-body protein components can alleviate miRNA-
mediated repression of gene expression.
Introduction
A PubMed search of “microRNA” or “miRNA” gives just over
1200 citations, all of them published since the turn of the mil-
lennium. These numerous publications have been concerned
primarily with using experimental and in silico approaches to
identify miRNAs and the target genes that they regulate in vari-
ous organisms, elucidating the biogenesis of miRNAs, and
investigating how the spectrum of miRNAs differs in different
tissues and disease states. Thus, by mid-2005, it was known that
the human genome encodes several hundred different
microRNAs (1) and that the pattern of miRNA expression is
often perturbed in disease states (2–4). Bioinformatic approaches
(1) further suggested that the mammalian miRNA repertoire
might collectively regulate several thousand genes, even though
only a handful of these predicted targets have been validated to
date. By contrast, very few of these studies addressed the mech-
anism of miRNA-mediated repression. Primary publications
Department of Biochemistry, University of Cambridge, 80 Tennis
Court Road, Cambridge CB2 1GA, UK.
*E-mail: rjj@mole.bio.cam.ac.uk (R.J.J.); nms@mole.bio.cam.ac.uk
(N.S.)
www.stke.org/cgi/content/full/2007/367/re1
and reviews written before that date generally proposed that
miRNAs (i) do not promote degradation of their target mRNAs,
but (ii) they do down-regulate target mRNA translation at some
stage after the initiation step.
Over the past 18 months, there has been a steady flow of
provocative reports on mechanisms of miRNA-mediated repres-
sion, and, far from consolidating and building upon these earlier
ideas, many of them directly contradict one or another of these
original assertions. The aim of this Review is to illuminate and
explain the controversies generated by these recent publications.
However, we have been unable to resolve many of the apparent
Downloaded from stke.sciencemag.org on February 26, 2009
contradictions, and our advice to the reader is to keep an open
mind to the possibility that there may be more than one mecha-
nism by which miRNAs effect posttranscriptional regulation of
gene expression.
This Review will therefore focus specifically on mechanisms,
with particular reference to metazoans. Lower eukaryotes are
mentioned only to the extent that they have contributed to our
understanding of mechanisms likely to operate in higher organ-
isms, and plants are ignored altogether, as the mechanisms of
plant miRNA biogenesis and action seem distinctly different.
Mechanisms of RNA interference (RNAi) by short interfering
RNAs (siRNAs) are also beyond the scope of this Review except
insofar as they relate directly to miRNA-mediated repression.
siRNAs Versus miRNAs: Definitions, Similarities, and
Differences
The generally accepted distinction between “miRNAs” and
“siRNAs” is that the former are ~21-residue RNAs derived
from longer RNAs that include a ~70-nucleotide (nt) imperfect-
ly base-paired hairpin segment that is the precursor of the ma-
ture miRNA, whereas siRNAs are of similar length but are de-
rived from longer, perfectly complementary double-stranded
RNA precursors of endogenous or exogenous origin (5). In both
cases, biogenesis proceeds through a staggered duplex interme-
diate, usually with ~19 base pairs (bp), and always with un-
paired 3′ extensions of 2 nt and a 5′-phosphate on each strand.
Especially in the case of mammalian cells, these ~21-nt RNAs
are often introduced by transfection of chemically synthesized
versions of this staggered duplex RNA, or alternatively as DNA
constructs that will give rise to such a duplex after transcription
and processing. One of the two strands, the one with the lower-
stability base-pairing at its 5′ end, is then assimilated into a
complex with several proteins, while the other, the so-called
passenger strand, is degraded [reviewed in (6)]. The resulting
RNA-protein complex is generally known as RISC (RNA-
induced silencing complex), or for miRNAs, it is sometimes
known as an miRNP (miRNA-protein complex).
For RNAi, the 21-residue siRNA, or its longer double-stranded
RNA precursor, is designed to be perfectly complementary to
part of the target mRNA sequence, which leads to endonucle-
olytic cleavage of the mRNA at the site of complementarity (6),
Page 1
 REVIEW

target sites are almost invariably located in the 3′-untranslated

A
region (3′-UTR) of a gene, often in multiple copies.
A A It is important to note that the formal distinction between an
mRNA 5’ ...AAGUUUUCAC AGCUAACA... 3’
CXCR4 3’ UUCAAAAGUG UCGAUUGUp 5’
siRNA and an miRNA is based solely on their different origins
A G and biogenesis, irrespective of the degree of complementarity to
G the target site and irrespective of whether the ~21-nt RNA pro-
G motes endonucleolytic cleavage of the target mRNA, or repres-
U A
U A sion of the target by some other mechanism. According to these
RL mRNA 5’ ...GCACAGCCUA CUACCUCA... 3’ definitions, a ~21-nt RNA introduced into cells by transfecting
let-7a 3’ UAUGUUGGAU GAUGGAGUp 5’ a chemically synthesized staggered duplex intermediate should
UUGG
be called an “siRNA,” even if its sequence has been designed to
B be precisely the same as that of a natural miRNA. This is a po-
tential source of confusion to those outside the immediate field,
...G U G U UA... and in order to avoid such difficulties, it might be preferable to
lin-28 5’ CAC GU UCUCAGG 3’ refer to such an RNA as a si(mi)RNA, to highlight the fact that
miR-125b 3’ GUG CA AGAGUCC 5’
A U U A C
C CUp although it is technically an siRNA, in fact it is one that is de-
UC signed to exactly mimic a bona fide miRNA.
Because the different outcomes depend on the degree of com-
plementarity between the short RNA and its target, it follows that
(C. e.) lin-14 5’ ...UC UAC CUCAGGGA C... 3’ a ~21-nt RNA designed to promote endonucleolytic cleavage of a

Downloaded from stke.sciencemag.org on February 26, 2009

3’ AG GUG GAGUCCCU Up 5’
lin-4 U A C
A particular mRNA with a perfectly complementary target site
A C
CU
U of this argument has been demonstrated for an siRNA designed
A U
5’ ...UUAUACAAC CUGCCUC...
(C. e.) lin-41
let-7 3’ GAUAUGUUG GAUGGAG
U A U Up
Fig. 1. Examples of the imperfect complementarity between vari-
ous miRNAs and their bulged target sites. Micro-RNA sequences
are in red, mRNA target sites in green. The two miRNA/mRNA
pairs in (A) have been widely used in studies of the mechanisms
of miRNA-mediated repression discussed here. The first is an
artificial system based on an siRNA designed for RNAi of
CXCR4 mRNA, but adapted to serve as a miRNA mimic through
the design of bulged target sites in the reporter mRNA 3′-UTR.
The second is the interaction of (endogenous) let-7a miRNA with
a reporter target site based on the predicted interaction of let-7a
with human lin-28 mRNA. The three examples in (B) are
endogenous miRNAs and their predicted (miR-125b/human
lin-28 mRNA) or validated (C. elegans lin-4 and let-7 miRNAs)
targets sites. Relevant references cited in this work are as
follows (note that these are not necessarily the publications that
first identified the miRNA/target site interaction): mRNA/CXCR4
(8, 11, 12, 33, 79); RL mRNA/let-7a (29); lin-28/miR-125b (47);
(C.e.) lin-14/lin-4 (22–25, 49); (C.e.) lin-41/let-7 (49).
followed by degradation of the two fragments by enzymes of the
normal mRNA decay pathways. Cleavage may not be entirely
abrogated by a very limited number of mismatches, depending
on their position, and so miRNAs that have a similar near-per-
fect complementarity to their target sites can also promote
degradation by this route (7). Generally, however, the interaction
of an miRNA with its predicted or validated target involves quite
extensive mismatched bulges, especially in the central region
(Fig. 1), and to a lesser extent in the miRNA 3′ end. This type of
interaction does not generally lead to mRNA degradation via en-
donucleolytic cleavage, but results in a decrease in protein ex-
pression that is usually (but not always) greater than the decrease
in mRNA abundance, suggesting that the main reason for down-
regulation is reduced efficiency of translation rather than in-
creased mRNA degradation. Validated and predicted miRNA
could also act as a translational repressor of another mRNA if
this has appropriate mismatched bulged target sites. The validity
3’ for RNAi of mRNA encoding the CXCR4 chemokine receptor
5’ (8). This siRNA has subsequently been widely used to study how
interaction of a ~21-nt RNA with imperfectly complementary 3′-
UTR sites results in repression of mRNA translation. For the rea-
sons explained above and because the sequence of this small
RNA shows no resemblance to any known natural miRNA, we
will refer to it as CXCR4 siRNA, but, except where otherwise
stated, it should be understood that it is always being used in con-
junction with mismatched target sites and is therefore serving as
a surrogate for a typical microRNA.
The mismatch between an miRNA and its target site can take
on different configurations, with a central unpaired bulge in ei-
ther the miRNA or mRNA strand, or both. Figure 1 shows both
natural miRNAs paired to their natural target, and the pairing of
the artificial CXCR4 siRNA with its laboratory-designed target
site. In the latter case, the exact configuration of the unpaired
bulge(s) affects the degree of repression, but this may not be ap-
plicable for other miRNA-mRNA pairs (9). Two extensive stud-
ies of how target site sequence and complementarity influence
repression by either natural Drosophila miRNAs overexpressed
in the wing imaginal disc or by the CXCR4 siRNA transfected
into HeLa cells are in very close agreement as to what is impor-
tant (10, 11). First and foremost, residues in the 5′ portion of the
miRNA (residues 2 to 8, the so-called seed) upstream of the dis-
continuity should be (almost) perfectly complementary to the
mRNA. The degree of repression is dependent on the stability of
the pairing in this region, with a rather sharp cut-off, and in ad-
dition, A-form geometry seems important because G-U pairs in
this region decrease repression to a greater extent than can be
explained purely on the grounds of mRNA stability. If the 5′ por-
tion of the miRNA is optimally base-paired, pairing between the
3′ portion and the mRNA target is not critical, but a high degree
of complementarity here can rescue repression when the pairing
of the miRNA “seed” is marginally suboptimal. In the CXCR4
system, the degree of repression is related to the number of tar-
get sites, at least up to six such sites, the maximum tested so far
(8, 12). In addition, different miRNAs can act in a combinatorial

www.stke.org/cgi/content/full/2007/367/re1 Page 2
 REVIEW
way, in the sense that cotransfection of two different siRNAs
with a reporter construct that had two 3′-UTR bulged target sites
for each of them resulted in a similar degree of repression, as
when there were four identical target sites for just one of the siRNAs
(11). Many endogenous mRNAs have just one or two predicted
sites for interaction with endogenous miRNAs, yet seem to be
quite efficiently repressed, which raises questions as to whether
the artificial systems such as CXCR4 may be intrinsically less
potent repressors in some way. This idea is given some support
by recent indications that the sequences flanking miRNA target
sites, or the context of such sites, can influence the biological
outcome of miRNA-mRNA interactions (13).
The contrasting outcomes that are dependent on whether
interaction of the ~21-nt small RNA with the mRNA target site
is perfectly complementary or has substantial bulged mismatches
are presumably due to different proteins being present in the
RISC and miRNP or, if the protein composition is really identi-
cal, then to the degree of complementarity influencing how
these proteins act. The protein compositions of RISC assembled
on perfectly complementary siRNAs of exogenous origin, and
the miRNPs formed with endogenous miRNAs, are often
described as similar or overlapping rather than absolutely iden-
tical. However, as the functional distinction between between a
perfectly complementary interaction and one with mismatches
is only apparent after the ~21-nt RNA has interacted with the
target site, any differences are likely to arise concomitantly
with, or shortly after, the establishment of this interaction.
Both types of complex include members of the Argonaute
family of proteins. Of the four mammalian Argonaute proteins,
only Ago2 is capable of siRNA-mediated endonucleolytic cleav-
age, with its ribonuclease H (RNase H)–like domain cutting the
phosphodiester bond in the target mRNA opposite the 10th and
11th residue in the siRNA (14, 15). Cocrystal structures show
that the 5′-phosphate on the siRNA is essential to place this
phosphodiester bond in the active site (16, 17). Such endonucle-
olytic cleavage would not be expected with most miRNA-mRNA
interactions because the A-form geometry would be severely dis-
rupted by the unpaired bulges in this region (Fig. 1).
Knockdown (via RNAi) of each Ago protein individually in
human embryonic kidney (HEK) 293 cells showed that Ago2
was the major contributor to miRNA-mediated repression of
target mRNA translation (18). Nevertheless, there is evidence
that Ago1, Ago3, and Ago4 are each capable of promoting this
repression to some extent (14, 19), even though they have no
known enzymatic activity. In contrast, the two Argonaute
proteins in Drosophila appear to have largely nonoverlapping
functions, with Ago1 involved mainly or exclusively in miRNA-
mediated repression, and Ago2 restricted to siRNA-mediated
endonucleolytic cleavage (20, 21).
Evidence That miRNAs Can Inhibit Translation at
Some Stage After Initiation
The pre-2005 hypothesis that miRNA-mediated repression of
target gene expression is due to inhibition of mRNA translation
at some stage after the initiation step was based on just two
studies of a single Caenorhabditis elegans miRNA, lin-4 miRNA
(the first microRNA to be discovered, long before the designa-
tion “microRNA” had been invented), and two larval mRNA
targets, lin-14 mRNA, which has up to seven potential 3′-UTR
sites for bulged interaction with lin-4 miRNA (22, 23), and
lin-28 mRNA. Expression of lin-4 miRNA commences late in
www.stke.org/cgi/content/full/2007/367/re1
the L1 larval stage, and by the late L2 or early L3 stage, the
amounts of LIN-14 and LIN-28 proteins are less than 10% of
that seen in mid-L1, but the abundance of lin-28 mRNA is es-
sentially unchanged, whereas that of lin-14 mRNA is about half
of that in the L1 stage (24, 25), with no appreciable difference
in poly(A) tail length (15 to 30 A residues).
In sucrose gradient analyses of polysomes at the late L2 or ear-
ly L3 stage, the repressed lin-14 and lin-28 mRNAs were found in
the same fractions as in a polysome distribution analysis of
mid-L1 larvae, where both would be efficiently translated (24,
25). When these polysomes from late L2 larvae were analyzed on
gradients containing EDTA, both mRNAs were found near the top
of the gradient, and on metrizamide gradients the lin-14 mRNA
was found in fractions of the same buoyant density as those con-
taining polysomes from L1 larvae (24, 25). Thus, by these two
commonly used criteria, lin-14 and lin-28 mRNAs appeared to be
genuinely in polysomes in late L2 and early L3 larvae, even
though their translation was repressed at these stages. These tests
do not, however, prove that the polysomes are dynamic and capa-
ble of elongation, an issue that can be examined by seeing whether
Downloaded from stke.sciencemag.org on February 26, 2009
the repressed mRNA moves into smaller polysomes on incubation
with puromycin, or with a specif ic inhibitor of initiation
(polysome runoff assays). These functional tests are best done in
intact cells, in case the translational repression might be relieved
by the mere act of preparing a cell-free extract, and should be
done over a relatively short time period (~5 min) so that differences
in the rate of reaction with puromycin or the rate of runoff can be
detected. However, puromycin and specific initiation inhibitors do
not penetrate C. elegans larvae, and so the best that could be done
was to isolate polysomes from early L3 larvae and add them to
reticulocyte lysate for polysome runoff, which showed that in a
rather long incubation period (45 min), the lin-28 mRNA moved
out of polysomes as effectively as did an mRNA that was not
repressed in the L3 larvae (25). The question of whether this
runoff produced any full-length LIN-28 protein was not examined.
The results described above are not a peculiarity of C. elegans
larval development. Similar results were obtained when DNA
transfections of 293T cells were used to study the repression of a
luciferase reporter with six bulged sites for interaction with
CXCR4 siRNA. Introduction of the CXCR4 siRNA reduced
overall luciferase expression by more than 95%, with relatively
little effect on target mRNA abundance (12). Likewise, the pres-
ence or absence of the CXCR4 siRNA had no influence on the
polysomal distribution of the target mRNA, nor did it affect the
movement of this mRNA into smaller polysomes on a brief
(3 min) incubation of the cells with puromycin. On a 5-min incu-
bation with a specific inhibitor of initiation, the target mRNA
moved into smaller polysomes, as expected, but the shift was dis-
tinctly greater under conditions of repression by CXCR4 siRNA.
This was interpreted as premature ribosome drop-off, or a failure
of processivity (12). However, if such drop-off was occurring
continuously, it is hard to see why the steady-state polysomal dis-
tribution of the target mRNA under repressed conditions would
be so similar to the unrepressed distribution, unless there was a
fortuitous increase in initiation frequency that precisely compen-
sated for the decrease in ribosome transit time. A plausible alter-
native explanation is that the target mRNA is slightly more sensitive
to the initiation inhibitor when it is repressed by the CXCR4
siRNA than it is under unrepressed conditions.
In this system, both cistrons of a dicistronic reporter with
either the hepatitis C virus (HCV) or cricket paralysis virus
Page 3
 REVIEW
(CrPV) IRES (internal ribosome entry site) driving downstream
cistron translation were susceptible to repression by CXCR4
siRNA, with the downstream cistron possibly even more sensi-
tive than the upstream one (12). Because initiation driven by the
CrPV IRES occurs by a highly unusual mechanism not requir-
ing any of the canonical translation initiation factors (26), this
result is consistent with the view that the repression mechanism
does not operate at the stage of initiation, but at some later step.
Cosedimentation of miRNAs with polysomes has been taken
as further evidence that repression occurs on target mRNAs
when they are actually in polysomes, and this is a valid
argument in cases where most of the cellular complement of
a particular miRNA is polysome-associated. However, if there
is only a minor fraction cosedimenting with polysomes, this
could possibly represent miRNAs associated with those mRNA
molecules that are not fully repressed. In late L2 stage C. elegans
larvae, some 10% of the total lin-4 miRNA cosedimented with
polysomes (24). Similarly, a small fraction of the miR-124a
present in a human neuronal cell line cosedimented with
polysomes in an EDTA-sensitive manner (27), and a more
substantial fraction of this and seven other miRNAs examined
was polysome-associated in rat forebrain neurons (28).
It is difficult to produce an explanation for an mRNA having
almost exactly the same polysomal distribution under condi-
tions of both strong miRNA-mediated repression and no repres-
sion. As Olsen and Ambros pointed out when this was first
discovered for lin-14 mRNA, inhibition of termination or an
inhibition of elongation that is less than 100% complete would
cause polysomes to increase in size (24). Conversely, an inhibi-
tion of elongation that is literally 100% complete will freeze
preexisting mRNA in polysomes, but any newly synthesized
mRNA will become associated with only a monomeric 80S
ribosome. In the context of miRNA regulation of translation,
neither of these scenarios has, to our knowledge, ever been
observed. On the basis of polysome dynamics alone, an miRNA-
mediated 90% decrease in protein output with no change in the
polysome distribution of the target mRNA would require a 90%
reduction in initiation frequency coupled with a balancing 90%
reduction in elongation rate, a somewhat implausible scenario.
These considerations led to the suggestion, likewise made by
Olsen and Ambros, that initiation and elongation rates on the
repressed mRNA may be normal, and the strong reduction in
protein product yield could be the result of a specific cotrans-
lational degradation of the nascent protein chain (24). For this
reason, the question of whether any full-length protein product
can be detected following polysome runoff, or in pulse-labeling
experiments, is important. In the Petersen et al. experiments,
pulse-labeling for 3 min failed to detect any product (12), even
though the reporter had an N-terminal epitope tag to facilitate
product detection, an observation that at face value is consistent
with the cotranslational degradation hypothesis.
However, there is an ingenious experiment that argues other-
wise. An mRNA encoding a protein with an N-terminal signal
sequence for targeting the product to the endoplasmic reticulum
(ER), and with three bulged sites for let-7a miRNA, was just as
susceptible to repression by endogenous HeLa cell let-7a as was
a luciferase reporter that encoded a cytoplasmic product (29).
Because the ER-targeted nascent protein chain is unlikely to be
accessible to any cytoplasmic proteolytic system, this result
argues against the cotranslational degradation hypothesis. It
should be noted, however, that in other experiments these
www.stke.org/cgi/content/full/2007/367/re1
authors found a polysome distribution of the repressed mRNA
suggestive of inhibited initiation (see below), and thus it remains
a possibility that cotranslational degradation of the nascent pro-
tein might occur specifically in those experimental conditions
where the repressed mRNA cosediments with large polysomes.
Evidence That miRNAs Can Inhibit Initiation Depen-
dent on eIF4E
The idea that miRNAs regulate mRNA translation at some stage
after the initiation step was challenged by the publication of two
papers about a year ago that point toward initiation as the regu-
lated step. The first, by Pillai et al., studied repression by
endogenous let-7a miRNA in HeLa cells, using luciferase
reporters with three bulged target sites in the 3′-UTR, with
sequestration of let-7a using an antisense 2′-O-methyl oligonu-
cleotide serving as a control (29). In DNA transfections,
luciferase expression was reduced by 80 to 90% at 48 hours
after transfection, but reporter mRNA abundance was reduced
by 20%. Polysome prof iles showed a distribution of the
repressed mRNA that was skewed toward the top of the gradi-
Downloaded from stke.sciencemag.org on February 26, 2009
ent, peaking in the 40S to disome region, strongly indicative of
inhibited initiation (29). A similar sucrose gradient distribution
of endogenous cationic amino acid transporter–1 mRNA was
seen under conditions where it was repressed by endogenous
miR-122 in Huh7 cells, which shows that this distribution is not
a peculiarity of overexpressed luciferase reporter targets (30).
In direct RNA transfections assayed at 6 hours after transfec-
tion, 7-methylguanosine (m7GpppG)–capped polyadenylated
mRNAs (where ppp represents a triphosphate moiety) that
would be translated by the scanning mechanism were suscepti-
ble to repression by let-7a, and were only slightly less sensitive
to this repression if the poly(A) tail was absent. By contrast, a
m7GpppG-capped polyadenylated reporter driven by the HCV
IRES was completely resistant, as were uncapped or adenosine
(ApppG)–capped nonpolyadenylated mRNAs dependent on the
encephalomyocarditis virus (EMCV) IRES (29). Because initia-
tion driven by the EMCV IRES requires all the canonical initia-
tion factors except eIF4E, the cap-binding factor (31), and the
HCV IRES functions independently of eIFs 4A, 4B, 4E, and 4G
(32), this result indicates that the repression mechanism affects
the function of eIF4E rather than any other factor.
These results are markedly similar to those in the second pa-
per, by Humphreys et al., who cotransfected HeLa cells with
CXCR4 siRNA and various luciferase reporter mRNAs, each
with four bulged target sites for this siRNA (33), the very same
reporters as used by the Sharp group (8, 11). At 16 hours after
transfection, the degree of CXCR4 siRNA-dependent repression
of a m7GpppG-capped polyadenylated mRNA was greater than
80% or about 50% if the poly(A) tail was absent. In addition,
50% repression was seen with an ApppG-capped polyadenylated
mRNA that would be translated by the scanning ribosome mech-
anism, and an ApppG-capped polyadenylated mRNA with an
EMCV IRES, whereas no repression was detectable with non-
polyadenylated versions of these RNAs or an ApppG-capped
nonpolyadenylated mRNA with the CrPV IRES (33).
The sensitivity of the m7GpppG-capped mRNA to repression,
contrasted with the lack of response of the nonpolyadenylated
mRNAs with the viral IRESs, is in complete agreement with Pil-
lai et al. and points to the same conclusion—an inhibition of
eIF4E function (29, 33). However, although the nonpolyadenylated
EMCV IRES construct was not repressed, adding a poly(A) tail
Page 4

increased its translation efficiency (in the absence of the CXCR4
siRNA) by a factor of ~2.5, but rendered it susceptible to a 50%
repression by this siRNA (33). In other words, the siRNA (al-
most) completely negated the stimulatory effect of the poly(A)
tail. This suggests that apart from any effect on eIF4E function,
the CXCR4 siRNA may also interfere with the “closed loop,” in
which poly(A) binding protein (PABP) bound to the poly(A) tail
interacts with the eIF4G component of the eIF4F complex (Fig.
2) bound either to the 5′ end of the mRNA (strongly in the case
of m7GpppG-capped RNAs and more weakly with an ApppG-
cap) or to an internal site in the EMCV IRES (34, 35). This
poly(A)-PABP–dependent “closed loop” enhances the efficiency
of initiation of both scanning-dependent mRNAs and mRNAs de-
pendent on the EMCV IRES, but the HCV IRES is unaffected by
the presence or absence of a poly(A) tail (36–39).
One can envisage two alternative ways by which the
CXCR4 siRNA could disrupt the “closed loop,” either via a
disruption of the protein-protein and protein-RNA interactions
necessary to maintain the loop, or by simply promoting dead-
enylation of the mRNA. Although no notable effect of the siRNA
A used in the other paper, may account for the higher repression
RNA helicase
(DEAD box)
eIF4A
PABP
elF4GI 570-582 722-876 1211-1411
130-160
elF4E
975-1065 eIF4A
Cap-binding
Sequence-independent
RNA-binding domain elF3
B capped dicistronic mRNA generated in the Petersen et al. proce-
eIF4E m7GpppG AUG
40S
4A
eIF
eIF4G elF3
BP UAA
PABP PA AA
AAA AA A
AAAAAAAAA
Fig. 2. Schematic diagram of (A) the eIF4F complex and (B) the
“closed loop” formed by the interaction of PABP bound to the
poly(A) with the eIF4G component of the eIF4F complex bound to
the 5′-cap. Mammalian cells have two eIF4Gs encoded by different
genes. Various isoforms of the more abundant species, eIF4GI,
arise from initiation at different AUG codons by leaking scanning.
The isoform shown, which is generally the most abundant, is the
second largest and lacks the 40 N-terminal amino acids of the
longest form. Apart from the direct interactions with eIF4A, eIF4E,
and PABP, eIF4G also interacts directly with eIF3, which itself
binds tightly to the 40S ribosomal subunit. Thus, there is an indi-
rect eIF4G/40S subunit interaction bridged by eIF3, and this is
thought to be instrumental in loading the 40S subunit on the mR-
NA and in facilitating ribosome scanning.
www.stke.org/cgi/content/full/2007/367/re1
REVIEW
on overall reporter mRNA stability was detected in these
experiments, the methods used would not have detected
changes in poly(A) tail length (33).
Can These Apparently Conflicting Results Be
Reconciled?
There are two, apparently unrelated, discrepancies between the
data described above and the results of Petersen et al.: (i) the dif-
ference in polysome distribution of the repressed mRNA, and
(ii) the issue of whether the viral IRESs are insensitive to repres-
sion (12, 29, 33). Although technical details of the polysome
analyses are not identical, there is no obvious difference that
might explain the disagreement. Both Pillai et al. and Petersen et
al. used DNA transfections for these experiments, albeit over
different time periods (48 and 20 hours, respectively) (12, 29).
One clear difference is that Pillai et al. relied on endogenous
let-7a to effect the repression (29), whereas Petersen et al. used
exogenous CXCR4 siRNA (12), which might well have resulted
in a higher intracellular concentration of the short RNA. This,
together with the fact that Petersen et al. had six bulged target
Downloaded from stke.sciencemag.org on February 26, 2009
sites in their reporter as compared to just three in the reporter
in their experiments, more than 95% as compared to 80 to 90%
in Pillai et al. One possibility is that at lower miRNA/mRNA
target ratios and lower repression ratios, inhibition is exclusively
or largely at the initiation step, whereas with increasing severity
1560
of repression a secondary effect on elongation rate or processivity
sets in.
As for the issue of whether IRES-dependent mRNAs are
insensitive to repression, in this case there are differences in pro-
cedure that deserve consideration as a possible explanation for the
40S subunit discrepancy: DNA transfections of constructs that would produce
a capped dicistronic mRNA (12), as opposed to RNA transfec-
tions of monocistronic IRES-containing RNAs (29, 33). The
dure would certainly bind the eIF4F complex (Fig. 2), and the
expected repression of the upstream cistron was indeed observed
(12). Is it possible that this repression of the upstream cistron,
however it may occur, could spread so as to also affect the down-
stream cistron? Unfortunately, this conjecture is undermined by
the results obtained by Pillai et al. with another type of dicistronic
mRNA (29), in which downstream cistron expression was driven,
not by a natural viral IRES, but by tandem phage λ Box B sites in
the intercistronic spacer, and coexpression of an eIF4G-λ N pep-
tide fusion protein, which would bind specifically to the Box B
sites (Fig. 3C). In DNA transfection assays, the presence of three
bulged target sites for let-7a in the 3′-UTR of this dicistronic con-
struct resulted in 65% repression of upstream cistron expression
with apparently no effect on the downstream cistron (29).
Different cell types and different miRNAs are unlikely to
explain the conflicting evidence implicating different mechanisms
of miRNA action. Petersen et al. and Humphreys et al. used pre-
cisely the same siRNA (CXCR4) and reporters with bulged target
sites of precisely the same sequence (though the number of sites
differed), yet came to completely different conclusions (12, 33).
Likewise, there appears to have been sufficient overlap of different
cell types used by each group to be able to eliminate this as the
explanation. Thus, there is no clear evidence pointing to one of the
two radically different outcomes being the right answer and the
other wrong, and the collective published data show no sizable
majority vote in favor of either explanation. If this and other
Page 5
 REVIEW
recent reviews (40) appear to have a bias in favor of initiation
being the regulated step, this is probably because there are numerous
precedents for regulation of eukaryotic initiation by protein-
mRNA interactions at 3′-UTR sites, but no precedents for cotrans-
lational degradation of the nascent protein, or for high-frequency
premature ribosome drop-off. However, a more detached and
unbiased position would take the view that miRNAs may be able
to mediate repression by either mechanism, but as these two out-
comes are so radically different, this would also imply that there
must be at least one unrecognized or uncontrolled variable that
determines which of the two putative alternative mechanisms is
predominant in particular experimental situations.
There are actually some precedents, dating from over 20 years
ago, for unexpected polysome association of near-silent mRNAs.
Heat shock of Drosophila cells generally leads to a complete dis-
appearance of polysomes (due to inhibition of initiation) followed
by the appearance of polysomes specifically translating heat-
shock mRNAs. However, in one report, some 30 to 50% of the
housekeeping mRNAs were still polysome- associated 1 hour af-
ter heat shock, even though the expression of the encoded
proteins was reduced by at least 90% (41). Similarly, prolonged
incubation of HeLa cells with amino acid analogs induces the
synthesis of stress proteins and shuts off synthesis of house-
keeping proteins, yet the mRNAs encoding these proteins remain
largely polysome-associated (42). Inhibition of initiation was
thought to be the immediate response to such stress, but it was
suggested that an additional elongation defect might set in
somewhat later. More recent confocal microscopy work reveals
an interesting parallel with miRNA-mediated repression in
that these near-silent housekeeping mRNAs would be largely
localized in cytoplasmic bodies known as stress granules
(43, 44), whereas mRNAs subject to miRNA-mediated repres-
sion localize to another type of cytoplasmic foci known as
P-bodies (see below).
Of course, the big difference between these situations and
miRNA-mediated repression is that the latter is selective for
(one or a few) particular mRNA species, whereas the stress con-
ditions affect the behavior of bulk mRNA. There are, however,
some examples in which a specific mRNA that appears to be
inactive is nevertheless still found in polysomes: thymidine
kinase mRNA in differentiating myoblasts (45), and nanos
mRNA in Drosophila embryos (46). The polysomes with
repressed nanos mRNA were sensitive to EDTA, they decreased
in size on prolonged incubation of the embryo extract with
puromycin, and polysome runoff resulted in release of most of
the nanos mRNA, yet no Nanos protein product could be detected
(46). These results are similar to the original observations on the
status of repressed lin-14 and lin-28 mRNAs in C. elegans
L2/L3 larvae (24, 25), and to the results of Petersen et al. (12).
miRNAs Can Also Promote Target mRNA Degradation
In almost all published studies, there is at least some miRNA-
dependent decrease in mRNA abundance, which has usually
been ignored, as it is often relatively small compared with the
inhibition of expression of the encoded protein. For example,
Petersen et al. observed a greater-than-95% decrease in reporter
expression, but only a 50% decrease in mRNA abundance (12).
Nevertheless, an explanation for this 50% decrease in mRNA
abundance is called for.
Some recent publications report larger miRNA-dependent
decreases in mRNA abundance, and emphasize mRNA degrada-
www.stke.org/cgi/content/full/2007/367/re1
A
N–Ago/GW182
Renilla luciferase
AAAA
5 Box B hairpins
bind N-peptide
Firefly luciferase as internal control
B
RL
( )
FL RL
Downloaded from stke.sciencemag.org on February 26, 2009
IRES
N–elF4E*/elF4G*
C
FL RL
2 to 6 Box B
Fig. 3. Examples of the IRES and tethered-function constructs
described here. (A) A reporter construct with five phage λ Box B
hairpins in its 3′-UTR is cotransfected with a construct designed
to express a fusion protein consisting of phage λ N-peptide fused
to a human or Drosophila Ago protein, or GW182 (and often with
an epitope tag spacer, which is not shown). The high-affinity in-
teraction between the N-peptide moiety and Box B hairpins teth-
ers the Ago or GW182 to the target mRNA. (B) Schematic dia-
gram illustrating the type of monocistronic and dicistronic mR-
NAs with viral IRESs discussed here. (C) A dicistronic mRNA in
which the downstream cistron expression is dependent on an ar-
tificial IRES system consisting of two or more phage λ Box B
hairpins. The DNA construct is cotransfected with constructs de-
signed to express a fusion protein consisting of phage λ N-pep-
tide fused to mammalian eIF4E or eIF4G mutants (eIF4E*,
eIF4G*) unable to function in cap-dependent initiation. The
eIF4G mutant is unable to interact with eIF4E but retains the
functionally important interactions with eIF3 and eIF4A (Fig. 2).
The eIF4E mutant cannot bind to 5′-caps, but retains the poten-
tial for interaction with wild-type endogenous eIF4G, and thus the
binding of the N-peptide/eIF4E* fusion protein to the Box B hair-
pins recruits endogenous eIF4G to the vicinity of the down-
stream cistron initiation site. Relevant references cited in this
work are as follows: (A) (19, 20, 29); (B) (12, 26, 29); (C) (29).
tion as an important aspect of miRNA-mediated repression of
gene expression. Microarray analyses of HeLa cell mRNAs
showed that expression of the brain-specific miR-124 or the
muscle-specific miR-1 decreased the abundance of different
subsets of 100 to 200 mRNAs in each case, specifically mRNAs
that are of low abundance in brain and muscle, respectively (47).
Page 6
 REVIEW
In addition, miRNA-dependent degradation of mRNAs that have
3′-UTR AU-rich destabilizing elements has been reported, even
though it is not the “seed” sequence of this miRNA (miR-16)
that is complementary to the AU-rich sequence (48). Even more
surprising is a recent report that in C. elegans larvae, lin-4 and
let-7 miRNAs promoted extensive degradation of lin-14, lin-28,
and lin-41 target mRNAs (49), notwithstanding the previous evi-
dence for repression of translation of these mRNAs (24, 25).
However, in none of these cases was there evidence that degra-
dation occurred by the endonucleolytic mechanism characteristic
of RNAi by perfectly complementary siRNAs (40).
Three recent papers, two concerned with mammalian sys-
tems and the other with Drosophila cells, also demonstrate
miRNA-mediated acceleration of target mRNA degradation, not
by an siRNA-like mechanism of endonucleolytic cleavage, but
rather, through the normal pathway of deadenylation, followed
by decapping and subsequent degradation of the body of the
mRNA by 5′→3′ exonuclease activity (18, 20, 50). The only
difference from the normal pathway is that the miRNP bound to
its 3′-UTR target sites can apparently act as a barrier to the
5′→3′ exonuclease, leading to accumulation of deadenylated
decay intermediates with 3′-end sequences (18, 49).
Thus, in Drosophila cells, target mRNA concentrations
could be restored by RNAi knockdown not only of Ago1 but al-
so of the deadenylating (Not1 and Ccr4) or decapping enzymes
(Dcp1/2), and in the last case deadenylated capped mRNAs ac-
cumulated (20). In mammalian cells with a target gene under
control of the c-fos promoter (to allow a transient transcriptional
pulse), an acceleration of deadenylation (and subsequent decay
of the mRNA body) dependent on miR-125b expression was
seen even with a reporter that had just a single bulged 3′-UTR
target site for this miRNA (50). The accelerated deadenylation
was not dependent on translation of the mRNA, because it was
not substantially affected by insertion of a stable hairpin in the
5′-UTR. Nor did premature decapping occur on the mRNAs
that had not yet been completely deadenylated (50).
Similarly, in zebrafish embryos, the expression of the miR-
430 family of microRNAs, which starts at ~5 hours after fertil-
ization, led to deadenylation and subsequent degradation of sev-
eral hundred maternal mRNAs with predicted bulged target
sites (and injected reporter mRNAs with such sites) at the stage
when zygotic transcription is activated. This serves to clear the
maternal mRNAs rapidly from the embryo and thus makes the
translation machinery available for the new transcripts. In mu-
tant embryos that do not express the miR-430 microRNAs,
deadenylation and degradation of the maternal transcripts still
occurred but were delayed, which led to abnormalities in brain
development (51).
Thus, mRNA degradation and inhibition of translation ap-
parently both contribute to the overall repression of gene ex-
pression by miRNAs, but the relative importance of each mech-
anism may vary according to the cell type and the stability or
configuration of the particular miRNA-mRNA pair (18, 20, 50).
This dual explanation for decreased expression is well illustrat-
ed by the data for reporters with the 3′-UTRs of three different
Drosophila genes: With the nerfin 3′-UTR, overall reporter pro-
tein output was reduced 90% by miR-9b expression, but mRNA
levels were reduced by only 20%, indicative of an effect pre-
dominantly on translation eff iciency; with the CG10011
3′-UTR, protein output and mRNA abundance were both
reduced by the same extent (80%) by miR-12 expression,
www.stke.org/cgi/content/full/2007/367/re1
suggesting almost no effect on translation efficiency per se; and
the Vha68-1 3′-UTR/miR-9b pair gave results intermediate be-
tween these two extremes (20).
A consequence of deadenylation is disruption of the “closed
loop” (Fig. 2), resulting in a severe reduction in translation effi-
ciency, which will fall effectively to zero after the subsequent
step of decapping. How much the low translation efficiency of
this capped deadenylated species contributes to the true repres-
sion of translation that is observed will depend on the lifetime
of this transient decay intermediate and its steady-state concen-
tration relative to that of the polyadenylated species. These
considerations are well illustrated by the three Drosophila
mRNAs mentioned above, where RNAi knockdown of decap-
ping enzymes caused the target mRNA abundance to increase
(by trapping the decay intermediates as capped deadenylated
transcripts), but the increase in reporter protein output was
smaller, so that the degree of true translational repression actu-
ally increased, in the sense that the ratio of protein output to the
amount of mRNA was reduced (20). However, because siRNA-
mediated depletion of deadenylating enzymes also increased
Downloaded from stke.sciencemag.org on February 26, 2009
target mRNA concentrations (to a lesser extent), but this time
with very little change in the degree of actual translational
repression (the ratio of protein output to mRNA abundance), it
seems likely that under normal conditions, translational repres-
sion is occurring mainly on polyadenylated mRNAs, with com-
paratively little contribution from the low translation efficiency
of the capped deadenylated mRNA decay intermediates. (The
situation may be somewhat different in RNA transfection
assays, where there is no ongoing production of fully
polyadenylated mRNAs.)
In mammalian cells expressing (from a CMV promoter) a
polyadenylated luciferase reporter mRNA with six bulged target
sites for miR-125b, coexpression of this miRNA reduced mRNA
abundance by about 70%, and luciferase protein expression by
more than 90%, indicative of a 65% decrease in the actual trans-
lation efficiency (50). When the poly(A) tail was replaced by the
3′-terminal stem-loop of histone H1.3, miR-125b coexpression
had little effect on mRNA abundance (just a 10% decrease), but
reduced translation efficiency by about 70%. The close corre-
spondence between the reduction in true translation efficiency in
these two experiments likewise suggests that the poor translata-
bility of capped deadenylated decay intermediates, which would
be present only in the former case, contributes little to the
observed translational repression. The issue of whether the
inhibition was exerted at initiation of translation or at some
subsequent step in protein synthesis was not addressed in this
study, but if future work indicates inhibition of initiation, it is
worth bearing in mind that, despite the absence of a poly(A) tail,
there is evidence that a “closed loop” (Fig. 2) can form with
histone mRNAs and that this stimulates their translation (52, 53).
miRNAs and Their mRNA Targets Accumulate in
P-Bodies
Another important development has been the discovery that
Argonaute proteins, miRNAs, and their target mRNAs accumu-
late in cytoplasmic foci usually known as P-bodies, or process-
ing bodies, which are detected by confocal microscopy and
immunostaining or the use of fusions between candidate P-body
proteins and a fluorescent protein (29, 54, 55). The protein
composition of these foci always includes enzymes that are
important in the normal pathway of mRNA degradation, but is
Page 7
 REVIEW
more complex in higher eukaryotes than in yeast, where it
includes the Dcp1/Dcp2 decapping complex, Xrn1 5′-3′ exonu-
clease, most of the Lsm proteins (Lsms 1 through 7, a family of
related RNA-binding proteins necessary for mRNA decay), the
Dhh1p putative RNA helicase (known as RCK or p54 in verte-
brates), and Pat1p, but not ribosomes, nor any of the eight trans-
lation initiation factors examined, nor components of the exo-
some complex of 3′→5′ exonucleases (56–58). Although Pat1p
homologs have not been tested, P-bodies in higher eukaryotes
seem to have the same proteins plus many additional compo-
nents, most of them absent from the yeast proteome, and
because budding yeast neither expresses siRNAs or miRNAs nor
shows any response to ~21-residue RNAs, at least some of these
additional components in higher eukaryotes could play a role in
the mechanisms of siRNA- and miRNA-mediated mRNA silenc-
ing. Interesting potential candidates include the following
proteins (59–62): GW182, a putative RNA-binding protein with
glycine-tryptophan repeats; eIF4E (but no other translation initi-
ation factors nor PABP nor ribosomes); and the eIF4E-binding
protein, 4E-T (but not 4E-BP1, another eIF4E binding protein).
The Drosophila 4E-T homolog, known as Cup, has been impli-
cated in the repression of specific maternal mRNAs (63–65),
whereas 4E-BP1 interaction with eIF4E results in global inhibi-
tion of initiation (66).
P-bodies are highly motile within the cell cytoplasm (44)
and fluctuate in size and number. In situations where P-bodies
seem to disappear as visible foci, it is not clear whether their
components become completely dispersed in the cytoplasm, or
whether “mini” P-bodies persist that are undetectable by confo-
cal microscopy. P-bodies not only contain (untranslated) mRNA
and especially mRNA decay intermediates, but they require this
RNA for their integrity (57, 60). Thus, the foci disappear in
yeast mutants with defective deadenylation, which would
reduce the level of decay intermediates, and increase in size and
number in mutants defective in downstream steps of the decay
process, such as mutants in the Dcp1/Dcp2 decapping complex
or the Xrn1 5′→3′ exonuclease (56, 57). Similar results are
seen in mammalian cells following RNAi knockdown of mRNA
degradation enzymes (60, 61), and in addition, P-bodies
decrease following depletion or knockdown of GW182, 4E-T,
Lsm1, and the RCK/p54 helicase (61, 67). Inhibition of transla-
tion elongation (e.g., by cycloheximide) causes the foci to dis-
appear, whereas inhibition of initiation results in an increase in
size and number (56–58, 60, 61), suggesting that an mRNA has
to be cleared of ribosomes before it can localize to P-bodies.
Thus, the P-body localization of the mRNA targets of miRNA-
mediated repression can be taken as additional, though circum-
stantial, evidence that it is initiation that is the inhibited step.
Irrespective of whether repression was effected by endoge-
nous let-7a miRNA or exogenous CXCR4 siRNA, a fraction of
the target mRNA was localized to foci resembling P-bodies (29,
54). However, Pillai et al. emphasize that the foci in which the
repressed mRNA and let-7 miRNA were localized did not
always coincide precisely with the Dcp-containing foci (29):
67% were perfectly overlapping with Dcp-1 foci, and 37% were
described as being adjacent. Furthermore, there were a few Dcp
foci apparently lacking the target mRNA, and vice versa, as
though there might be two subtypes of foci that might fuse
occasionally or exchange mRNA. In a similar vein, adjacent
location and possible fusion between P-bodies and stress gran-
ules has also been noted (43, 44, 68), leading to the suggestion
www.stke.org/cgi/content/full/2007/367/re1
that after storage and sorting in stress granules, some mRNAs
may pass to P-bodies for degradation.
Entry of an mRNA into Dcp1/Dcp2 foci does not inevitably
lead to its degradation, as shown by the fact that glucose starva-
tion of yeast, which results in a global inhibition of initiation,
drives mRNAs into P-bodies, but on refeeding they exit the foci
and become (re)assimilated into active polysomes (58, 69).
Reversibility of miRNA-dependent repression and P-body
localization has been demonstrated for the cationic amino acid
transporter (CAT-1) mRNA, which has three or four [depending
on which poly(A) addition site was used] bulged 3′-UTR sites
for interaction with miR-122 (30). In Huh7 cells, the endoge-
nous miR-122 reduced CAT-1 protein levels by about 65%
under fed conditions (as judged by the results of introducing an
antisense oligonucleotide to this miRNA), and both the CAT-1
mRNA and the miR-122 could be detected in P-bodies. On
amino acid starvation there was a rapid increase in the amount
of CAT-1 protein, with no immediate change in mRNA abun-
dance; rather, the preexisting CAT-1 mRNA left the P-bodies
and was mobilized into polysomes. An unexpected additional
Downloaded from stke.sciencemag.org on February 26, 2009
finding was that both the activation of translation and the mobi-
lization from P-bodies required a particular AU-rich element
(ARE) in the 3′-UTR and the presence of the ARE-binding pro-
tein, HuR (30). Together, these observations conf irm that
miRNA-mediated repression and P-body localization of the
target mRNA are potentially reversible processes. P-bodies can
be regarded as a repository for untranslated mRNA, some of
which may subsequently be degraded, whereas some may be
returned to translation. An unresolved question is whether the
translational activation of CAT-1 mRNA is accompanied by
complete dissociation of the miRNP from the target mRNA, or
whether it is just a case of the repressive action being overrid-
den in some way without dissociation.
It was estimated that some 20% of let-7a miRNA and 20%
of its target mRNA in HeLa cells were localized in foci, under
conditions in which the target was repressed by at least 80%
(29). The results of biochemical fractionation of cell extracts
prepared by lysis with digitonin suggest that the repression of
the rest of the target mRNA occurred in “mini” P-bodies smaller
than the threshold size detectable by confocal microscopy,
rather than uniformly and completely dispersed throughout the
cytoplasm. Centrifugation of these extracts at 14,000g pelleted
almost all the target mRNA (but not the tubulin mRNA exam-
ined as an unrepressed control), most of the let-7a miRNA (in
the case of HeLa cells), and all of the endogenous miR-122
(Huh7 cells), together with all of the cellular complement of the
P-body marker proteins, Xrn1 and Dcp1 (29, 30).
It seems intuitive that it is the inhibition of translation rather
than the mere binding of the miRNP to the target mRNA that
drives the repressed mRNA into P-bodies. Thus, it would be
reasonable to assume that in the experiments where mRNAs
with the EMCV, HCV, or CrPV IRESs were resistant to repres-
sion (29, 33), an miRNP-mRNA complex still forms on these
mRNAs, yet presumably they do not get localized to P-bodies.
It would be intriguing to know what happens as regards P-body
accumulation with a dicistronic mRNA (Fig. 3C), in which the
upstream cistron is sensitive to miRNA-mediated repression but
the downstream cistron is insensitive (29).
Another case where it seems most unlikely that the target
RNA enters P-bodies is the interaction of miR-122 with HCV
RNA, which occurred at a 5′-proximal site upstream of the
Page 8
 REVIEW
IRES and was necessary for the replication of the viral RNA,
without having any direct effect on its translation (70). Here
again, there is no reason to suppose that an miRNP complex
with Ago proteins does not form on the viral RNA, yet it seems
most unlikely that this results in the miRNP/viral RNA complex
accumulating in P-bodies.
siRNA-Mediated Knockdown and Tethered Function
Approaches to Identify the Essential Mediators of
miRNA-Dependent Repression of Translation
Approaches toward elucidating the mechanism of repression in-
clude (i) identifying which proteins interact with Argonaute pro-
teins, (ii) testing whether miRNA-mediated repression is relieved
by siRNA-mediated knockdown or depletion of these and other
likely candidate proteins, and (iii) tethered function assays (Fig. 3).
Proteins that interact with mammalian Ago1 or Ago2 pro-
teins in an RNase-insensitive manner, and colocalize in P-bodies,
include the Dcp1-Dcp2 decapping complex, GW182, and
RCK/p54 helicase, which is the vertebrate homolog of yeast
Dhh1p (54, 67, 71). The interactions with Dcp1/2 and GW182
still occurred with Ago mutants deficient in RNA binding that
did not localize to P-bodies (54, 67), and the interaction with
p54 still occurred in cells depleted of Lsm1 (a protein found in
P-bodies and implicated in RNA processing), which dispersed
the foci (71). Again, it is not clear whether these interacting
proteins were completely dispersed in these situations. The
Drosophila GW182 homolog binds to and localizes with Ago1,
but not Ago2 (20), which functions mainly or exclusively in the
siRNA-mediated endonucleolytic pathway.
When Drosophila cells were depleted of these various pro-
teins, relief of miRNA-mediated repression was seen with
knockdown of Ago1, GW182, and Dcp1/Dcp2, in descending
order of efficacy (20, 72). Depletion of Xrn1 had no effect, nor
did Ago2 knockdown, though this did relieve siRNA-mediated
endonucleolytic cleavage of mRNA with a perfectly comple-
mentary target site (72). In mammalian cells, repression of re-
porters with either four or six bulged target sites for CXCR4
siRNA was likewise relieved by GW182 depletion and, to a
slightly lesser extent, by knockdown of p54 helicase or Dcp2,
but not by Xrn1 knockdown (67, 71). Depletion of Lsm1, which
caused dispersal of P-bodies, had no effect on the degree of re-
pression, implying that the usual localization in visible foci is a
consequence of repression rather than a prerequisite for it (71).
In the case of a reporter with a single site perfectly complemen-
tary to the introduced siRNA, depletion of GW182 again gave
some relief, though not as much as Ago2 knockdown, and
Dcp1/2 or Xrn1 depletion had no effect (67, 71).
In view of the indications that miRNA-mediated repression in
some way affects eIF4E function (29, 33), it would be interesting
to test the outcome of 4E-T knockdown. Another candidate
worth testing would be the higher-eukaryote homolog of yeast
Pat1p, which has been shown to act as a global translational re-
pressor together with Dhh1p (the yeast p54 homolog), in an ad-
ditive rather than overlapping way (69). In strains deleted for
both, the normal response to glucose starvation, which includes
the breakdown of polysomes, cessation of translation, and mobi-
lization of all mRNAs into P-bodies, no longer occurred (69).
Finally, tethered function assays have shown that coexpres-
sion of Ago-λ N-peptide fusions and a reporter with tandem
phage λ Box B sequences in its 3′-UTR (Fig. 3A) caused the re-
porter to behave as if it were repressed by microRNAs. Its
www.stke.org/cgi/content/full/2007/367/re1
translation was specifically repressed at the initiation step by up
to 80%, dependent on the number of Box B sites, but indepen-
dent of their position in the 3′-UTR, with no appreciable de-
crease in mRNA abundance (19, 29). Curiously, although re-
pression promoted by tethering Ago bypasses the requirement
for an miRNA, repression and P-body localization did not occur
with tethered Ago PAZ domain mutants that are defective in
binding the miRNA/mRNA complex (67).
Tethering GW182 to an mRNA in the same way mimicked
miRNA-mediated repression in Drosophila cells, causing a
75% decrease in the abundance of the reporter mRNA and a
greater than 90% overall reduction of luciferase output, imply-
ing a 75% inhibition of translation per se (20). The same effects
of tethered GW182 were seen even in cells depleted of Ago1.
Thus, tethering the relevant Argonaute protein can bypass the
miRNA requirement for repression, and tethering GW182
bypasses the Ago requirement, at least in Drosophila cells,
suggesting that GW182 acts downstream of Ago proteins.
As for p54, tethering it to the 3′-UTR of a reporter mRNA
specifically repressed reporter translation in Xenopus oocytes,
Downloaded from stke.sciencemag.org on February 26, 2009
implying that p54 plays an important role in the represssion of
maternal mRNA in these oocytes (73, 74). The Drosophila p54
homolog, Me31B, has also been implicated in the repression of
certain maternal mRNAs in early embryos, and in this capacity
it appears to act as a corepressor working in conjunction with
Cup, the Drosophila homolog of mammalian 4E-T (75, 76). It
is not yet known if tethering 4E-T/Cup to the 3′-UTR also
represses translation, and, if so, whether it or p54 acts down-
stream of GW182 or vice versa.
Recapitulation of miRNA-Mediated Repression in Vitro
Many of the uncertainties about the exact mechanism of transla-
tional repression might be resolved if repression could be reca-
pitulated in an in vitro system, as has been achieved for siRNA-
mediated endonucleolytic cleavage (77, 78). Until recently, no
such in vitro recapitulation of miRNA-mediated repression had
been reported, although we presumed that it had been attempted
in many laboratories without success. However, success has
recently been achieved with the same siRNA (CXCR4) and a
luciferase mRNA with the same bulged target sites as used by
Petersen et al. (12) and Humphreys et al. (33). The essential
trick was to pre-anneal the miRNA to the target mRNA before
addition to the rabbit reticulocyte lysate (RRL) system (79).
Although this requirement for pre-annealing means that the
miRNP-mRNA complex was not assembled by the normal route
that is thought to occur in intact cells, the outcome had all the
hallmarks of miRNA-mediated regulation established in vivo:
increased repression with increasing number of target sites; a
strict requirement for a 5′-phosphate on the miRNA; require-
ment for perfectly complementary contiguous base-pairing of
residues 2 to 8 of the miRNA, coupled with a much less strin-
gent requirement for contiguous base-pairing of the 3′ end; no
repression if the 3′-end of the 21-nt miRNA was extended by 10
residues perfectly complementary to the target mRNA; and no
repression with a 2′-O-methyl derivative of the 21-nt mRNA.
Although the commercial RRL system used is one that appears
to have been optimized for translation of uncapped mRNA,
repression was only seen if the mRNA was both capped and had
a standard-length poly(A) tail (200 A residues), but uncapped
mRNAs were repressed if the tail was extended to unphysiolog-
ical lengths (2000 residues).
Page 9
 REVIEW

Maximal repression (70% reduction) of luciferase output A Accelerated

was seen at short incubation times (10 min), and the time- deadenylation
course data implied that the CXCR4 siRNA increased the lag
time before the luciferase activity first appeared, after which the Inhibition
actual rate of increase was not very different. Although these of initiation
miRNP
kinetics could imply a delay in the first initiation events, or
alternatively a slower rate of elongation, it could also be that the m7GpppG AUG UAA AAAAAAAA
repression mechanism is rather labile over time, a suggestion
which would be consistent with the finding that repression by A
CXCR4 siRNA was more sensitive to freeze-thawing of the Defective B
RRL than was the basal translation activity (79). Although the elongation
data do not allow an unambiguous identification of which step
of translation is being inhibited, the requirement for a cap and a
poly(A) tail strongly suggests that it is initiation rather than
elongation that is affected.
If the reporter mRNA had a single 3′-UTR site perfectly A. Cotranslational degradation of nascent protein
B. Premature ribosome drop-off
complementary to the CXCR4 siRNA, it appeared to be exten- C. Reduced elongation rate (and initiation frequency)
sively degraded in this system, implying that despite the
abnormal pathway, a competent RISC with an active Ago2 can
assemble in RRL, although confirmation of this requires a B

Downloaded from stke.sciencemag.org on February 26, 2009

demonstration that the reporter was cleaved in the expected site.
AAA
A Possible Model? Pat1
AA
AA
elF4E AA
We propose a speculative model to account for the available Ccr4 A
findings (Fig. 4). Figure 4A depicts all the different ways by

AA
Not1

AA
which miRNAs have been proposed to regulate their target 4E-T p54

A
AAA
mRNAs, other than the rare cases where metazoan miRNAs

have near-perfect complementarity with an mRNA target site
and therefore promote endonucleolytic cleavage (7). There are
two reasons why we are unable to enlarge upon the possible
mechanisms of perturbation of elongation that result in the
repressed mRNA being found in polysomes of approximately
the same size as when it is unrepressed. First, it is not yet at all
clear whether this is due to cotranslational cleavage of the
nascent protein chain, or to premature ribosome drop-off
(coupled with a compensating increase in initiation frequency),
or to a reduced elongation rate coupled with a similar decrease
in the rate of initiation. Second, none of the components of
miRNPs, and none of the proteins that have been found to inter-
act with Ago proteins, is thought to interact with, or have any
impact upon an elongating ribosome..
However, we are able to elaborate the model in respect of
miRNA-mediated regulation of translation initiation and
miRNA-dependent acceleration of mRNA degradation (Fig.
4B). In this scheme, binding of an Ago protein at the site of
miRNA interaction with its bulged target sequence leads to
recruitment of GW182, RCK/p54, the Dcp1/Dcp2 decapping
complex, 4E-T, and the higher eukaryote homolog of yeast
Pat1p (although there is as yet no evidence for or against
recruitment of these last two components). With the assis-
tance of p54, 4E-T interacts with eIF4E bound to the 5′-cap,
to form an inhibitory “closed loop” similar to the postulated
mechanism for repression of maternal mRNAs in Drosophila
embryos by the combined action of Cup and Me31B, the
Drosophila homologs of 4E-T and p54, respectively (63–65,
75). An additional assumption is that the bound Ago protein
recruits, either directly or indirectly, the Not1 and Ccr4 dead-
enylation factors, which results in accelerated shortening of
the poly(A) tail and consequently accelerated mRNA degra-
dation. This model cannot explain why repressed mRNAs and
miRNAs sometimes end up in polysomes (12, 24, 25), nor
AA
GW182 Dcp1/2
Ago
Fig. 4. Ultraspeculative model for how miRNAs may be able to
regulate gene expression at the posttranscriptional level. (A) A
schematic depiction of the different mechanisms proposed for
the regulation of target mRNA function by miRNAs. Three
alternative ways in which miRNA-mediated repression might
perturb translation elongation are listed, with the green sphere
representing an unknown ribosome-associated protease
degrading the nascent protein chain cotranslationally. (B) Dia-
gram showing how the binding of an Ago protein to the site of
miRNA/mRNA interaction might lead to inhibition of translation
initiation and an enhanced rate of deadenylation. (See text for
an explanation.) For convenience, the diagram depicts an
mRNA with only a single target site for the miRNA. With multi-
ple sites there is the possibility that protein interactions at one
site could be directed mainly toward inhibition of initiation, and
those at another site toward deadenylation, or even toward
aberrant elongation.
can it explain how and why translation dependent on the
HCV and CrPV IRESs is susceptible to repression in some
experiments (12). But apart from these problems, the model
is consistent with the other results discussed here, subject to
the minor caveat that the weak repression seen with ApppG-
capped polyadenylated mRNAs (with or without an EMCV
IRES) in the RNA transfection assays of Humphreys et al.
would have to be explained as a secondary consequence of
deadenylation rather than a direct result of any inhibition of
initiation (33).

www.stke.org/cgi/content/full/2007/367/re1 Page 10
 REVIEW
Concluding Remarks
A comment found with increasing frequency in the literature is
that there may well be more than one mechanism by which miRNAs
repress expression of their target genes. There is clear evidence
for this in respect to deadenylation-dependent degradation and
true translational repression, and it does not stretch credulity too
far to imagine that the miRNA/mRNA interaction could lead to
both these outcomes. What does defy the imagination, on the
other hand, is how the same miRNA-mRNA target pairs can
show inhibition of translation initiation in one system, but inhi-
bition at some step after initiation in another. We embarked on
this Review with the hope of finding an explanation for this ap-
parent contradiction. We have failed in this aim, but some point-
ers to future directions have emerged on the route to this failure.
First, more attention should be paid than hitherto to the
poly(A) tail length of the target mRNA. Moreover, target
mRNA abundance should in future be assessed by methods
(e.g., Northern blotting) that, unlike RNase protection or
reverse transcription–polymerase chain reaction amplification
(RT-PCR) of a short internal segment, can distinguish full-
length mRNA from degradation intermediates. Second, it would
be interesting to know whether depletion of critical P-body
components such as GW182 and RCK/p54 also relieves the re-
pression in those systems that lead to polysomal accumulation
of the repressed mRNA. Third, it would be useful if those who
find that target mRNAs with an HCV, CrPV, or EMCV IRES
are insensitive to repression in RNA transfection assays were to
confirm this result with dicistronic mRNAs, and also in DNA
transfections of mono- and dicistronic constructs. For this pur-
pose, it would also be worthwhile replacing the HCV IRES with
the phylogenetically related, but much more potent, IRES from
classical swine fever virus (80) as a counter to the argument
(40) that translation dependent on some viral IRESs may be
unresponsive to miRNAs simply because these IRESs are
inefficient. Fourth, because initiation on all the viral IRESs
used to date is not only independent of eIF4E but also does not
involve any ribosome scanning, it seems important to extend the
range of IRESs studied to those that require eIF4E and those
where there is ribosomal scanning, so as to confirm that it is
eIF4E function rather than scanning that is inhibited. Fifth,
further use of depletions can be expected to identify more can-
didate proteins required for repression, and tethered function
assays coupled with knockdowns may reveal the hierarchy in
which these function. Sixth, further development of the cell-free
systems (79) can be expected to illuminate the proximal mecha-
nism of repression, though it would be somewhat surprising if
such systems were able to shed light on the mechanisms of
deadenylation and target mRNA accumulation in P-bodies.
Finally, in order to facilitate reconciliation of the biochemical
fractionation data with the confocal microscopy results, it
would be extremely useful to know what happens to P-bodies,
and where they and their constituent proteins partition, in the pro-
cedures used by different laboratories to prepare cytoplasmic ex-
tracts for fractionation by sucrose density gradient centrifugation.
Note added in proof. There have been two very recent publi-
cations on these controversial issues. Both agree that repression
occurs while the target mRNA is polysome-associated, implying
that the repression must be exerted at some stage after initiation.
In sucrose gradient analyses of HeLa cell polysomes, virtually
the whole cellular complement of three representative miRNAs
(let-7a, miR-16, and miR-21) cosedimented with polysomes in
www.stke.org/cgi/content/full/2007/367/re1
an EDTA-sensitive manner (81). This polysomal association was
dependent on base-pairing between the miRNA and mRNA,
because the let-7a miRNA was found at the top of the gradient if
the cells had been transfected with a 2′-O-methyl oligonu-
cleotide antisense to this miRNA. Moreover, the polysomes were
dynamic by the criteria that the miRNAs were found nearer the
top of the gradient if the cells had been incubated with
puromycin or with initiation inhibitors before lysis (81).
In the other study, HeLa cells were transfected with a DNA
construct with a thymidine kinase promoter that had the Renilla
luciferase coding region upstream of the C. elegans lin-41
3′-UTR, which has two validated target sites for let-7a miRNA
(82). As compared with a control lacking these two sites,
luciferase expression was reduced by 90%, with a decrease in
target mRNA abundance of 30%, implying an ~80% reduction
in true translation efficiency. About half of the target mRNA,
as well as half the endogenous let-7a miRNA and AGO pro-
teins, cosedimented with large polysomes (with the other half
found at the top of the gradients), but they sedimented nearer
the top of the gradient following incubation with puromycin or
Downloaded from stke.sciencemag.org on February 26, 2009
under conditions of inhibited initiation (82). These results were
taken as evidence that repression was occurring at some stage
after initiation, and the authors specifically favor the cotransla-
tional nascent protein degradation hypothesis, because, like
Petersen et al. (12), they were unable to detect any nascent
peptide product under repressed conditions, even though the
reporter had an N-terminal epitope tag to facilitate detection of
such peptide products.
References and Notes
1. B. P. Lewis, C. B. Burge, D. P. Bartel, Conserved seed pairing, often
flanked by adenosines, indicates that thousands of human genes are
microRNA targets. Cell 120, 15–20 (2005).
2. J. Lu, G. Getz, E. A. Miska, E. Alvarez-Saavedra, J. Lamb, D. Peck, A.
Sweet-Cordero, B. L. Ebert, R. H. Mak, A. A. Ferrando, J. R. Downing, T.
Jacks, H. R. Horvitz, T. R. Golub, MicroRNA expression profiles classify
human cancers. Nature 435, 834–838 (2005).
3. L. He, J. M. Thomson, M. T. Hemann, E. Hernando-Monge, D. Mu, S.
Goodson, S. Powers, C. Cordon-Cardo, S. W. Lowe, G. J. Hannon, S. M.
Hammond, A microRNA polycistron as a potential human oncogene.
Nature 435, 828–833 (2005).
4. I. Alvarez-Garcia, E. A. Miska, MicroRNA functions in animal develop-
ment and human disease. Development 132, 4653–4662 (2005).
5. V. Ambros, B. Bartel, D. P. Bartel, C. B. Burge, J. C. Carrington, X. Chen,
G. Dreyfuss, S. R. Eddy, S. Griffiths-Jones, M. Marshall, M. Matzke, G.
Ruvkun, T. Tuschl, A uniform system for microRNA annotation. RNA 9,
277–279 (2003).
6. Y. Tomari, P. D. Zamore, Perspective: Machines for RNAi. Genes Dev.
19, 517–529 (2005).
7. S. Yekta, I. H. Shih, D. P. Bartel, MicroRNA-directed cleavage of HOXB8
mRNA. Science 304, 594–596 (2004).
8. J. G. Doench, C. P. Petersen, P. A. Sharp, siRNAs can function as
miRNAs. Genes Dev. 17, 438–442 (2003).
9. M. Kiriakidou, P. T. Nelson, A. Kouranov, P. Fitziev, C. Bouyioukos, Z. Moure-
latos, A. Hatzigeorgiou, A combined computational-experimental approach
predicts human microRNA targets. Genes Dev. 18, 1165–1178 (2004).
10. J. Brennecke, A. Stark, R. B. Russell, S. M. Cohen, Principles of
microRNA-target recognition. PLoS Biol. 3, e85 (2005).
11. J. G. Doench, P. A. Sharp, Specificity of microRNA target selection in
translational repression. Genes Dev. 18, 504–511 (2004).
12. C. P. Petersen, M.-E. Bordeleau, J. Pelletier, P. A. Sharp, Short RNAs re-
press translation after initiation in mammalian cells. Mol. Cell 21,
533–542 (2006).
13. D. Didiano, O. Hobert, Perfect seed pairing is not a generally reliable
predictor for miRNA-target interactions. Nat. Struct. Mol. Biol. 13,
849–851 (2006).
14. J. Liu, M. A. Carmell, F. V. Rivas, C. G. Marsden, J. M. Thomson, J. J.
Song, S. M. Hammond, L. Joshua-Tor, G. J. Hannon, Argonaute2 is the
catalytic engine of mammalian RNAi. Science 305, 1437–1441 (2004).
15. G. Meister, M. Landthaler, A. Patkaniowska, Y. Dorsett, G. Teng, T.
Page 11
 REVIEW
Tuschl, Human Argonaute2 mediates RNA cleavage targeted by
miRNAs and siRNAs. Mol. Cell 15, 185–197 (2004).
16. J. S. Parker, S. M. Roe, D. Barford, Crystal structure of a PIWI protein
suggests mechanisms for siRNA recognition and slicer activity. EMBO
J. 23, 4727–4737 (2004).
17. J. B. Ma, K. Ye, D. J. Patel, Structural basis for overhang-specific small inter-
fering RNA recognition by the PAZ domain. Nature 429, 318–322 (2004).
18. D. Schmitter, J. Filkowski, A. Sewer, R. Pillai, E. Oakeley, M. Zavolan, P.
Svoboda, W. Filipowicz, Effects of Dicer and Argonaute down-regulation
on mRNA levels in human HEK293 cells. Nucleic Acids Res. 34,
4801–4815 (2006).
19. R. S. Pillai, C. G. Artus, W. Filipowicz, Tethering of human Ago proteins
to mRNA mimics the miRNA-mediated repression of protein synthesis.
RNA 10, 1518–1525 (2004).
20. I. Behm-Ansmant, J. Rehwinkel, T. Doerks, A. Stark, P. Bork, E. Izaur-
ralde, mRNA degradation by miRNAs and GW182 requires both
CCR4:NOT deadenylase and DCP1:DCP2 decapping complexes.
Genes Dev. 20, 1885–1898 (2006).
21. J. Rehwinkel, P. Natalin, A. Stark, J. Brennecke, S. M. Cohen, E. Izaur-
ralde, Genome-wide analysis of mRNAs regulated by Drosha and Arg-
onaute proteins in Drosophila melanogaster. Mol. Cell. Biol. 26,
2965–2975 (2006).
22. R. C. Lee, R. L. Feinbaum, V. Ambros, The C. elegans heterochronic
gene lin-4 encodes small RNAs with antisense complementarity to
lin-14. Cell 75, 843–854 (1993).
23. B. Wightman, I. Ha, G. Ruvkun, Posttranscriptional regulation of the
heterochronic gene lin-14 by lin-4 mediates temporal pattern formation
in C. elegans. Cell 75, 855–862 (1993).
24. P. H. Olsen, V. Ambros, The lin-4 regulatory RNA controls developmen-
tal timing in Caenorhabditis elegans by blocking LIN-14 protein synthe-
sis after the initiation of translation. Dev. Biol. 216, 671–680 (1999).
25. K. Seggerson, L. Tang, E. G. Moss, Two genetic circuits repress the
Caenorhabditis elegans heterochronic gene lin-28 after translation
initiation. Dev. Biol. 243, 215–225 (2002).
26. J. E. Wilson, T. V. Pestova, C. U. Hellen, P. Sarnow, Initiation of protein
synthesis from the A site of the ribosome. Cell 102, 511–520 (2000).
27. P. T. Nelson, A. G. Hatzigeorgiou, Z. Mourelatos, miRNP:mRNA association
in polyribosomes in a human neuronal cell line. RNA 10, 387–394 (2004).
28. J. Kim, A. Krichevsky, Y. Grad, G. D. Hayes, K. S. Kosik, G. M. Church,
G. Ruvkun, Identification of many microRNAs that copurify with polyri-
bosomes in mammalian neurons. Proc. Natl. Acad. Sci. U.S.A. 101,
360–365 (2004).
29. R. S. Pillai, S. N. Bhattacharyya, C. G. Artus, T. Zoller, N. Cougot, E.
Basyuk, E. Bertrand, W. Filipowicz, Inhibition of translational initiation by
let-7 microRNA in human cells. Science 309, 1573–1576 (2005).
30. S. N. Bhattacharyya, R. Habermacher, U. Martine, E. I. Closs, W.
Filipowicz, Relief of microRNA-mediated translational repression in
human cells subjected to stress. Cell 125, 1111–1124 (2006).
31. T. V. Pestova, C. U. Hellen, I. N. Shatsky, Canonical eukaryotic initiation
factors determine initiation of translation by internal ribosomal entry.
Mol. Cell. Biol. 16, 6859–6869 (1996).
32. T. V. Pestova, I. N. Shatsky, S. P. Fletcher, R. J. Jackson, C. U. Hellen, A
prokaryotic-like mode of cytoplasmic eukaryotic ribosome binding to the
initiation codon during internal translation initiation of hepatitis C and
classical swine fever virus RNAs. Genes Dev. 12, 67–83 (1998).
33. D. T. Humphreys, B. J. Westman, D. I. Martin, T. Preiss, MicroRNAs
control translation initiation by inhibiting eukaryotic initiation factor
4E/cap and poly(A) tail function. Proc. Natl. Acad. Sci. U.S.A. 102,
16961–16966 (2005).
34. V. G. Kolupaeva, T. V. Pestova, C. U. Hellen, I. N. Shatsky, Translation
eukaryotic initiation factor 4G recognizes a specific structural element
within the internal ribosome entry site of encephalomyocarditis virus
RNA. J. Biol. Chem. 273, 18599–18604 (1998).
35. I. B. Lomakin, C. U. Hellen, T. V. Pestova, Physical association of
eukaryotic initiation factor 4G (eIF4G) with eIF4A strongly enhances
binding of eIF4G to the internal ribosomal entry site of encephalomy-
ocarditis virus and is required for internal initiation of translation. Mol.
Cell. Biol. 20, 6019–6029 (2000).
36. A. B. Sachs, P. Sarnow, M. W. Hentze, Starting at the beginning, middle
and end: Translation initiation in eukaryotes. Cell 89, 831–838 (1997).
37. H. Imataka, A. Gradi, N. Sonenberg, A newly identified N-terminal
amino acid sequence of human eIF4G binds poly(A)-binding protein
and functions in poly(A)-dependent translation. EMBO J. 17,
7480–7489 (1998).
38. G. Bergamini, T. Preiss, M. W. Hentze, Picornavirus IRESes and the
poly(A) tail jointly promote cap-independent translation in a mammalian
cell-free system. RNA 6, 1781–1790 (2000).
39. Y. M. Michel, A. M. Borman, S. Paulous, K. M. Kean, Eukaryotic initiation
factor 4G-poly(A) binding protein interaction is required for poly(A) tail-
mediated stimulation of picornavirus internal ribosome entry segment-
www.stke.org/cgi/content/full/2007/367/re1
driven translation but not for X-mediated stimulation of hepatitis C virus
translation. Mol. Cell. Biol. 21, 4097–4109 (2001).
40. M. A. Valencia-Sanchez, J. Liu, G. J. Hannon, R. Parker, Control of
translation and mRNA degradation by miRNAs and siRNAs. Genes
Dev. 20, 515–524 (2006).
41. D. G. Ballinger, M. L. Pardue, The control of protein synthesis during
heat shock in Drosophila cells involves altered polypeptide elongation
rates. Cell 33, 103–113 (1983).
42. G. P. Thomas, M. B. Mathews, Alterations of transcription and transla-
tion in HeLa cells exposed to amino acid analogs. Mol. Cell. Biol. 4,
1063–1072 (1984).
43. N. Kedersha, G. Stoecklin, M. Ayodele, P. Yacono, J. Lykke-Andersen,
M. J. Fitzler, D. Scheuner, R. J. Kaufman, D. E. Golan, P. Anderson,
Stress granules and processing bodies are dynamically linked sites of
mRNP remodeling. J. Cell Biol. 169, 871–884 (2005).
44. P. Anderson, N. Kedersha, RNA granules. J. Cell Biol. 172, 803–808 (2006).
45. M. K. Gross, G. F. Merrill, Thymidine kinase synthesis is repressed in
nonreplicating muscle cells by a translational mechanism that does not
affect the polysomal distribution of thymidine kinase mRNA. Proc. Natl.
Acad. Sci. U.S.A. 86, 4987–4991 (1989).
46. I. E. Clark, D. Wyckoff, E. R. Gavis, Synthesis of the posterior determi-
nant Nanos is spatially restricted by a novel cotranslational regulatory
mechanism. Curr. Biol. 10, 1311–1314 (2000).
47. L. P. Lim, N. C. Lau, P. Garrett-Engele, A. Grimson, J. M. Schelter, J.
Castle, D. P. Bartel, P. S. Linsley, J. M. Johnson, Microarray analysis
shows that some microRNAs downregulate large numbers of target
Downloaded from stke.sciencemag.org on February 26, 2009
mRNAs. Nature 433, 769–773 (2005).
48. Q. Jing, S. Huang, S. Guth, T. Zarubin, A. Motoyama, J. Chen, F. Di
Padova, S. C. Lin, H. Gram, J. Han, Involvement of microRNA in AU-rich
element-mediated mRNA instability. Cell 120, 623–634 (2005).
49. S. Bagga, J. Bracht, S. Hunter, K. Massirer, J. Holtz, R. Eachus, A. E.
Pasquinelli, Regulation by let-7 and lin-4 miRNAs results in target
mRNA degradation. Cell 122, 553–563 (2005).
50. L. Wu, J. Fan, J. G. Belasco, MicroRNAs direct rapid deadenylation of
mRNA. Proc. Natl. Acad. Sci. U.S.A. 103, 4034–4039 (2006).
51. A. J. Giraldez, Y. Mishima, J. Rihel, R. J. Grocock, S. Van Dongen, K.
Inoue, A. J. Enright, A. F. Schier, Zebrafish miR-430 promotes dead-
enylation and clearance of maternal mRNAs. Science 312, 75–79
(2006).
52. J. Ling, S. J. Morley, V. M. Pain, W. F. Marzluff, D. R. Gallie, The histone
3′-terminal stem-loop-binding protein enhances translation through a
functional and physical interaction with eukaryotic initiation factor 4G
(eIF4G) and eIF3. Mol. Cell. Biol. 22, 7853–7867 (2002).
53. B. Gorgoni, S. Andrews, A. Schaller, D. Schumperli, N. K. Gray, B.
Muller, The stem-loop binding protein stimulates histone translation at
an early step in the initiation pathway. RNA 11, 1030–1042 (2005).
54. J. Liu, M. A. Valencia-Sanchez, G. J. Hannon, R. Parker, MicroRNA-
dependent localization of targeted mRNAs to mammalian P-bodies.
Nat. Cell Biol. 7, 719–723 (2005).
55. G. L. Sen, H. M. Blau, Argonaute 2/RISC resides in sites of mammalian mR-
NA decay known as cytoplasmic bodies. Nat. Cell Biol. 7, 633–636 (2005).
56. U. Sheth, R. Parker, Decapping and decay of messenger RNA occur in
cytoplasmic processing bodies. Science 300, 805–808 (2003).
57. D. Teixeira, U. Sheth, M. A. Valencia-Sanchez, M. Brengues, R. Parker,
Processing bodies require RNA for assembly and contain nontranslat-
ing mRNAs. RNA 11, 371–382 (2005).
58. M. Brengues, D. Teixeira, R. Parker, Movement of eukaryotic mRNAs
between polysomes and cytoplasmic processing bodies. Science 310,
486–489 (2005).
59. T. Eystathioy, A. Jakymiw, E. K. Chan, B. Seraphin, N. Cougot, M. J. Frit-
zler, The GW182 protein colocalizes with mRNA degradation associat-
ed proteins hDcp1 and hLSm4 in cytoplasmic GW bodies. RNA 9,
1171–1173 (2003).
60. N. Cougot, S. Babajko, B. Seraphin, Cytoplasmic foci are sites of mRNA
decay in human cells. J. Cell Biol. 165, 31–40 (2004).
61. M. A. Andrei, D. Ingelfinger, R. Heintzmann, T. Achsel, R. Rivera-Pomar,
R. Lührmann, A role for eIF4E and eIF4E-transporter in targeting
mRNPs to mammalian processing bodies. RNA 11, 717–727 (2005).
62. M. A. Ferraiuolo, S. Basak, J. Dostie, E. L. Murray, D. R. Schoenberg,
N. Sonenberg, A role for the eIF4E-binding protein in P-body formation
and mRNA decay. J. Cell Biol. 170, 913–924 (2005).
63. J. E. Wilhelm, M. Hilton, Q. Amos, W. Henzel, Cup is an eIF4E binding
protein required for both the translational repression of oskar and the
recruitment of Barentsz. J. Cell Biol. 163, 1197–1204 (2003).
64. A. Nakamura, K. Sato, K. Hanyu-Nakamura, Drosophila Cup is an
eIF4E binding protein that associates with Bruno and regulates oskar
mRNA translation in oogenesis. Dev. Cell 6, 69–78 (2004).
65. M. R. Nelson, A. M. Leidal, C. A. Smibert, Drosophila Cup is an eIF4E-
binding protein that functions in Smaug-mediated translational
repression. EMBO J. 23, 150–159 (2004).
Page 12
 REVIEW
66. A. Pause, G. Belsham, A.-C. Gingras, O. Donzé, T.-A. Lin, J. Lawrence Jr.,
N. Sonenberg, Insulin-dependent stimulation of protein synthesis by phos-
phorylation of a regulator of 5′-cap function. Nature 371, 762–767 (1994).
67. J. Liu, F. V. Rivas, J. Wohlschlegel, J. R. Yates III, R. Parker, G. J. Han-
non, A role for the P-body component GW182 in microRNA function.
Nat. Cell Biol. 7, 1261–1266 (2005).
68. A. Wilczynska, C. Aigueperse, M. Kress, F. Dautry, D. Weil, The transla-
tional regulator CPEB1 provides a link between Dcp1 bodies and stress
granules. J. Cell Sci. 118, 981–992 (2005).
69. J. Coller, R. Parker, General translational repression by activators of
mRNA decapping. Cell 122, 875–886 (2005).
70. C. L. Jopling, M. Yi, A. M. Lancaster, S. M. Lemon, P. Sarnow, Modula-
tion of hepatitis C virus RNA abundance by a liver-specific microRNA.
Science 309, 1577–1581 (2005).
71. C. Y. Chu, T. M. Rana, Translation repression in human cells by microR-
NA-induced gene silencing requires RCK/p54. PLoS Biol. 4, e210 (2006).
72. J. Rehwinkel, I. Behm-Ansmant, D. Gatfield, E. Izaurralde, A crucial role
for GW182 and the DCP1:DCP2 decapping complex in miRNA-mediat-
ed gene silencing. RNA 11, 1640–1647 (2005).
73. N. Minshall, G. Thom, N. Standart, A conserved role of a DEAD box he-
licase in mRNA masking. RNA 7, 1728–1742 (2001).
74. N. Minshall, N. Standart, The active form of Xp54 RNA helicase in
translational repression is an RNA-mediated oligomer. Nucleic Acids
Res. 32, 1325–1334 (2004).
75. A. Nakamura, R. Amikura, K. Hanyu, S. Kobayashi, Me31B silences
translation of oocyte-localising RNAs through the formation of cytoplas-
mic RNP complex during Drosophila oogenesis. Development 128,
3233–3242 (2001).
www.stke.org/cgi/content/full/2007/367/re1
76. M. Chekulaeva, M. W. Hentze, A. Ephrussi, Bruno acts as a dual repres-
sor of oskar translation, promoting mRNA oligomerization and forma-
tion of silencing particles. Cell 124, 521–533 (2006).
77. J. Martinez, A. Patkaniowska, H. Urlaub, R. Luhrmann, T. Tuschl, Single-
stranded antisense siRNAs guide target RNA cleavage in RNAi. Cell
110, 563–574 (2002).
78. J. Martinez, T. Tuschl, RISC is a 5′ phosphomonoester-producing RNA
endonuclease. Genes Dev. 18, 975–980 (2004).
79. B. Wang, T. M. Love, M. E. Call, J. G. Doench, C. D. Novina, Recapitula-
tion of short RNA-directed translational gene silencing in vitro. Mol. Cell
22, 553–560 (2006).
80. S. P. Fletcher, I. K. Ali, A. Kaminski, P. Digard, R. J. Jackson, The influ-
ence of viral coding sequences on pestivirus IRES activity reveals
fur ther parallels with translation initiation in prokaryotes. RNA 8,
1558–1571 (2002).
81. P. A. Maroney, Y. Yu, J. Fisher, T. W. Nilsen, Evidence that microRNAs
are associated with translating messenger RNAs in human cells. Nat.
Struct. Mol. Biol. 13, 1102–1107 (2006).
82. S. Nottrott, M. J. Simard, J. D. Richter, Human let-7a miRNA blocks
protein production on actively translating polyribosomes. Nat. Struct.
Mol. Biol. 13, 1108–1114 (2006).
83. We thank V. Ambros, J. Belasco, E. Izaurralde, C. Jopling, E. Miska, C.
Novina, R. Parker, T. Preiss, and D. Scadden for helpful comments on
the draft versions of this article.
Downloaded from stke.sciencemag.org on February 26, 2009
Citation: R. J. Jackson, N. Standart, How do microRNAs regulate gene
expression? Sci. STKE 2007, re1 (2007).
Page 13